Charmed Kubernetes on Equinix Metal

As with any cloud supported by Juju, Charmed Kubernetes can be deployed and used on Equinix Metal. This document provides some extra information and an overlay to help get the most out of this cloud. For instructions on installing Juju itself, please see the latest Juju documentation.

Before installing

Equinix Metal has been added to the clouds Juju automatically knows about. To check, you can run the command:

juju list-clouds --all

If equinix does not appear in the list, your local Juju install probably just needs to refresh its list of clouds. Run:

juju update-public-clouds

You should also add your credentials for this cloud. Use the interactive command:

juju add-credential equinix

...and follow the prompts to enter the information required (including the project id, and your auth token).

Installing

To deploy Charmed Kubernetes on Equinix Metal, it is also recommended to deploy some storage and to use Calico for networking. You can deploy and configure Charmed kubernetes any way you like, but this example bundle will help you get started.

You can download the example bundle here

It can then be installed with the command:

juju deploy ./equinix-bundle.yaml

When the deployment has settled, remember to fetch the configuration file!

juju scp --proxy kubernetes-master/0:config ~/.kube/config

You can check the status by running:

juju status

At this point, there will be error messages on the workers as the pods will not run until the Cloud Controller Manager has been run.

Post install

To use Kubernetes on Equinix Metal, you should now set up the Equinix Cloud Controller Manager.

While the deployment is in progress no pods will be able to spun up on the Kubernetes due to taints being set on each node. The taints will be removed once the Cloud Controller Manager (CCM) is enabled and the nodes are registered with the cloud control plane.

First, a Kubernetes secret has to be created, defining the variables for the CCM:

kubectl create -f - <<EOY
apiVersion: v1
kind: Secret
metadata:
  name: metal-cloud-config
  namespace: kube-system
stringData:
  cloud-sa.json: |
    {
    "apiKey": "<Metal API key>",
    "projectID": "<Metal Project ID>",
    "loadbalancer": "kube-vip://"
    }
EOY

The next steps are to confirm the version of the CCM to use:

export CCM_VERSION=3.2.2
kubectl apply -f https://github.com/equinix/cloud-provider-equinix-metal/releases/download/v${CCM_VERSION}/deployment.yaml

...enable premissions for the kube-vip loadbalancer:

kubectl apply -f https://kube-vip.io/manifests/rbac.yaml

... and deploy:

kubectl apply -f - <<EOY
apiVersion: apps/v1
kind: DaemonSet
metadata:
  creationTimestamp: null
  name: kube-vip-ds
  namespace: kube-system
spec:
  selector:
    matchLabels:
      name: kube-vip-ds
  template:
    metadata:
      creationTimestamp: null
      labels:
        name: kube-vip-ds
    spec:
      containers:
      - args:
        - manager
        env:
        - name: vip_arp
          value: "false"
        - name: vip_interface
          value: lo
        - name: port
          value: "6443"
        - name: vip_cidr
          value: "32"
        - name: svc_enable
          value: "true"
        - name: vip_startleader
          value: "false"
        - name: vip_addpeerstolb
          value: "true"
        - name: annotation
          value: metal.equinix.com
        - name: bgp_enable
          value: "true"
        - name: bgp_routerid
        - name: bgp_as
          value: "65000"
        - name: bgp_peeraddress
        - name: bgp_peerpass
        - name: bgp_peeras
          value: "65000"
        - name: vip_address
        image: plndr/kube-vip:v0.3.6
        imagePullPolicy: Always
        name: kube-vip
        resources: {}
        securityContext:
          capabilities:
            add:
            - NET_ADMIN
            - NET_RAW
            - SYS_TIME
      hostNetwork: true
      serviceAccountName: kube-vip
  updateStrategy: {}
status:
  currentNumberScheduled: 0
  desiredNumberScheduled: 0
  numberMisscheduled: 0
  numberReady: 0
EOY

Note: in some Equinix Metal facilities it is required to define a static route on each Kubernetes Worker node to allow the traffic to the workloads exposed via the Load Balancer to go via proper gateway:

juju exec --application kubernetes-worker 'GATEWAY_IP=$(curl https://metadata.platformequinix.com/metadata | jq -r ".network.addresses[] | select(.public == false) | .gateway"); sudo ip route add 169.254.255.1 via $GATEWAY_IP; sudo ip route add 169.254.255.2 via $GATEWAY_IP'

Using load balancers

With the cloud load balancer capabilities enabled, actions which invoke a loadbalancer in Kubernetes will automatically trigger creation of the ElasticIP in the Metal cloud and associate it with the KubeVIP service, simultaneously adjusting BGP tables in the cloud and forward the traffic to Kubernetes nodes. This can be demonstrated with a simple application. Here we will create a simple application and scale it to five pods:

kubectl create deployment hello-world --image=gcr.io/google-samples/node-hello:1.0
kubectl scale deployment hello-world --replicas=5

You can verify that the application and replicas have been created with:

kubectl get deployments hello-world

Which should return output similar to:

NAME              READY   UP-TO-DATE   AVAILABLE   AGE
hello-world      5/5               5                            5             2m38s

To create a LoadBalancer, the application should now be exposed as a service:

kubectl expose deployment hello-world --type=LoadBalancer --name=hello --port 8080

To check that the service is running correctly:

kubectl get service hello

...which should return output similar to:

NAME    TYPE           CLUSTER-IP       EXTERNAL-IP   PORT(S)          AGE
hello   LoadBalancer   10.152.183.136   202.49.242.3  8080:32662/TCP   2m

You can see that the External IP is now in front of the five endpoints of the example deployment. You can test the ingress address:

curl  http://202.49.242.3:8080
Hello Kubernetes!

We appreciate your feedback on the documentation. You can edit this page or file a bug here.